But why would they invest in more factories if they also think it's a temporary hickup?
Then it's just the same capacity, but without huge buyers. Still the prices won't come down...
If they built new factories now, they would just lose money to an investment that would not pay off.
(Of course under the premise that AI collapses or is saturated at some point. If that doesn't hold true then ex falso quodlibet!)
> But why would they invest in more factories if they also think it's a temporary hickup?
Because they have orderbooks 2 years (at least) into the future so know what demand is there - and they are demanding deposits for future orders.
It's easy to see if this is true. Look for news on new factories opening:
Micron: https://finance.yahoo.com/technology/ai/articles/micron-mu-p...
Samsung: https://www.kedglobal.com/korean-chipmakers/newsView/ked2026... (note this is doubling Samsung's memory production)
SK Hynix: https://finance.yahoo.com/sectors/technology/articles/sk-hyn...
Yes, 2 years.
How long does building a factory take?
If the demand grows with their production they can sell more units at the same price.
If demand goes down by a certain percentage, they sell more for less + they lost the investment into new factories.
It all is based on IFs and about personality, about "optimism" vs "pessimism"
I for one think that the AI bubble will "burst" at some point and I think that then there will be a lot of hardware to go by.
Time will be the judge of my abilities to replace the Oracle of Delphi.